azureml-featurestore

What it is and what it does

The azureml-featurestore package is the Python SDK for Azure ML's managed feature store, designed to work alongside azure-ai-ml. It lets you define feature sets with Spark-based transformations, list and retrieve feature specifications, and run offline feature retrieval using point-in-time joins—a key pattern in ML pipelines where you need historical feature values aligned to specific timestamps.

The package supports multiple feature definition approaches: a Domain Specific Language (DSL) for declarative transformations, user-defined functions (UDF), or no transformation. It can load from materialized stores, handle temporal joins with lookback windows, and materialize data between offline and online stores. Runtime dependencies include azure-ai-ml (the parent SDK), mltable (for table abstractions), jinja2 (for templating), marshmallow (for serialization), and pandas (for data handling).

Use it for:

• Define and manage feature sets in Spark with custom transformations for ML model training pipelines.
• Retrieve historical feature values at specific points in time for training dataset generation.
• Materialize computed features from offline storage into online Redis cache for batch scoring.
• List and inspect feature specifications already defined in your Azure ML Feature Store.
• Build feature engineering workflows using DSL syntax without writing custom transformation code.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

Provides Python SDK access to Azure ML Feature Store for developing feature sets, managing feature specifications, and running offline feature retrieval with point-in-time joins.

Yes, if you are already invested in Azure ML and need a managed feature store with offline retrieval and point-in-time join capabilities. The package is production-stable, has low install friction, and carries permissive licensing. However, the 189-day gap since last release and aging maintenance status suggest slower iteration; verify that the current feature set (including online store maturity) meets your timeline and requirements before committing to it for new projects.

Install

azureml-featurestore on PyPI

pip install azureml-featurestore

uv add azureml-featurestore

poetry add azureml-featurestore

Installing azureml-featurestore

Before you install

Low install friction with a pure-Python wheel. Maintenance status is aging—last release was 189 days ago—but the package is marked Production/Stable and supports current Python versions (3.8–3.12).

License in practice

MIT License (permissive) allows broad use, modification, and distribution with minimal restrictions.

Quickstart

pip install azureml-featurestore

from azureml.featurestore import FeatureStoreClient
from azure.ai.ml import MLClient

# Initialize clients
ml_client = MLClient.from_config()
fs_client = FeatureStoreClient(ml_client=ml_client)

Requires Azure credentials and an existing Azure ML workspace configured via azure-ai-ml.

Verify before relying

• Whether offline feature retrieval performance scales to production workload sizes.
• Current state of online feature store support and its maturity beyond public preview.
• Whether DSL feature definition syntax is stable or subject to breaking changes.

Package facts

Evidence: azureml_featurestore-1.2.2-py3-none-any.whl

Keywords: AzureMachineLearning, azure, feature store

Tags

More Artificial Intelligence packages